We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
VentureBeat AI·20d ago·5 min read
Agentic orchestration: Enterprise AI organizations have a deployment problem, not a platform problem — and most are calling chatbots agents
TL;DR
Enterprises are struggling with deploying AI agents, primarily using chatbot frameworks instead of more advanced orchestration. Despite the popularity of platforms like Anthropic’s Claude, many organizations face challenges in achieving reliable multi-step execution.
✦ Why It Matters
Engineers should explore advanced orchestration frameworks to enhance AI agent capabilities beyond basic chatbot functionalities.
Key Takeaways
- Investigate advanced orchestration tools for AI deployment.
- Evaluate the cost implications of token usage in AI applications.
- Consider hybrid control planes to avoid vendor lock-in.
- Focus on multi-step execution capabilities in AI agents.
Full Summary
Agent orchestration in enterprises is increasingly centered around model-provider platforms, with Anthropic’s Claude being the most favored due to its robust underlying model. However, a survey of 101 enterprises reveals that most implementations are limited to chatbot wrappers, lacking the sophisticated control and orchestration expected.
Companies are opting for hybrid control planes to prevent vendor lock-in, which complicates the deployment landscape. Furthermore, real-time fiscal control over token usage—essential for managing costs—is still rare.
These findings indicate a significant gap between the potential of AI agents and their current practical applications. For engineers and researchers, this suggests a need to focus on developing more advanced orchestration capabilities and cost management strategies.